VModal Android SDK: a UI-free Kotlin client for hosted video search and streaming uploads
Video Search SDK for Android Kotlin. Integrate in any video app
At a glance
- What is it?
- The VModal Android SDK is a Kotlin client for a hosted multimodal video and image search service. It handles the gateway, request models, signed upload streams, progress and cancellation, and leaves every screen to your app. Using it requires an API key from the service.
- Who is it for?
- The VModal Android SDK suits Android teams that want semantic search over uploaded video and images without building their own indexing service, and that are happy to depend on a hosted API with a key. It is a poor fit if you need on-device search or a ready-made user interface, since the SDK ships no screens and calls a remote service.
- Can I use it commercially?
- Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 43 days ago.
- What is it written in?
- Mainly Kotlin, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- v-modal/vmodal_sdk_android544 ★ · -44
| Repository | 2026-09-17 | 2026-11-15 |
|---|---|---|
| v-modal/vmodal_sdk_android | 588 | 544 |
-44 Stars in the last 59 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
The SDK owns the transport and your app owns every screen
V-Modal AI for Android describes itself as a way to give an Android app a visual memory for video and images: you upload video, then find moments by image-semantic meaning or imagery. The division of work is stated clearly. Your app owns the screens and lifecycle, while the SDK handles the gateway, request models, response parsing, signed upload streams, progress and cancellation.
It is a UI-free Kotlin client. The documentation says the SDK publishes no navigation, screens, themes, accessibility policy or design system, and that demo UI remains application-owned. That suits teams with an existing design system, in Compose or in classic Views, but it means a working search screen is something you build, using the example applications as a starting point. The pitch also names Kotlin, Compose, Views and coroutines as the tools you already know, which tells you the API is designed around standard Android patterns instead of a new framework.
One immutable scope binds project, collection and stream
For new content flows, the SDK ties upload, search, asset, index and deletion calls to one immutable project, collection and stream. The sample configures a client with a project ID and an API key your app loaded, then scopes it:
import com.vmodal.sdk.VModal
val content = VModal.configure(
projectId = "food_app",
apiKey = apiKeyLoadedByYourApp,
).scope(
collectionName = "user_123",
streamName = "uploads",
)
val results = content.search("the cyclist crossing the bridge at sunset")The collection name in the sample is a user identifier, which shows one way to organise data per user. Because the scope is immutable, switching accounts means creating a new scoped client, and the cookbook lists account-switch cleanup as a topic you need to handle. Index organization examples cover global, per-user, multi-stream and catalog layouts.
Uploads stream from content URIs and can be cancelled for real
Uploads work directly from the system photo picker, because the SDK supports streaming content:// URIs, so you do not copy a large video into app storage first. Uploads are exposed as a cold Flow, with a callback-style UploadHandle kept for compatibility, which gives you a cancel action that actually cancels.
For work that must survive beyond one screen, the SDK pairs a CoroutineWorker with a cancellation-aware upload Flow, so WorkManager can continue an upload after the user leaves the screen. Progress is reported through the same flow. A separate upload and WorkManager guide, a coroutines guide and the integration cookbook cover the details, including typed failures, which let you separate a network error from a rejected request.
Credentials stay with your app, and a hosted service sits behind every call
Authentication uses app-owned runtime credentials, with no login UI imposed, so the SDK slots into existing authentication and dependency injection. A guide on managing API keys safely is part of the documentation, and the sample passes the key in from a variable your app loads, not as a literal.
The dependency on a service is the main thing to understand. Search and indexing happen on VModal's hosted platform, which is why an API key is needed, and the README offers a free beta API key through a contact page. Nothing in the visible README describes on-device indexing, so assume that videos you upload leave the phone and that search needs a network connection. Review the provider's data handling before uploading user content.
Search is semantic, and the sample query is a sentence
The feature table lists semantic video and image search, with the example Find the red car entering the parking lot, and a fixed image-semantic search for finding a visual moment by meaning. In the code sample the query is a plain sentence about a cyclist crossing a bridge at sunset, passed to the search method on the scoped client.
That is a different input from keyword search. You describe what appears in the footage, and the service returns matching moments. Because relevance depends on the hosted model, test your own queries against your own content, and design the screen to handle an empty result gracefully. The README gives no accuracy figures or latency numbers, so there is nothing to compare against. Place a loading state on the screen as well, because a search that goes to a remote service takes a network round trip each time.
The agent prompt refers to a directory that the top level does not show
A collapsible section offers a prompt to paste into a coding agent. It asks the agent to clone the repository and then run the install script from a subdirectory:
git clone https://github.com/v-modal/vmodal_sdk_android
cd vmodal_sdk_androidThe prompt then says to change into uinterface/sdk_android, run bash install.sh with install and then check, and to use the checked-in Gradle wrapper without creating another environment. But the repository's top level contains api, assets, docs, examples, gradle, src and a settings file, with no uinterface directory. The prompt appears to come from a larger source tree, and the repository has a .vmodal-public-source marker file. If the agent fails at that step, it is following a path that does not exist in this checkout, and the Gradle wrapper at the root is the thing to use.
Six example projects, no releases and a CI badge that points elsewhere
The examples directory has numbered projects: starter examples for one API operation at a time, an upload and search app, a full search application validated stage by stage, index organization examples, a Framebase Compose application for local street-video search, and a JSON metadata example. The docs directory includes the integration cookbook, a technical introduction, a coroutines guide and a search app guide.
The repository has no GitHub releases, so there are no tagged versions to pin from here, and installation is deferred to the technical introduction. The last push was on 2026-10-03 and the license is Apache-2.0. The continuous integration badge links to a workflow in a different repository, vmx_api, which fits the idea that this repository is a published copy of source maintained elsewhere. An osv-scanner configuration is included for dependency scanning.
Editorial conclusion
The VModal Android SDK suits Android teams that want semantic search over uploaded video and images without building their own indexing service, and that are happy to depend on a hosted API with a key. It is a poor fit if you need on-device search or a ready-made user interface, since the SDK ships no screens and calls a remote service. Before building, get an API key, confirm the repository layout against the install prompt, and note that the repository has no tagged releases.
Frequently asked questions
What is the purpose of Android SDK?
For this project, the V-Modal Android SDK is a Kotlin client that lets an Android app upload video and search it by image-semantic meaning through a hosted service.
What does modal do?
V-Modal AI provides multimodal video and image search. The Android SDK adds semantic search and streaming uploads to a Kotlin app while leaving the screens to your code.
How to use modal?
Configure the client with VModal.configure using a project ID and an API key, call scope with a collection and stream name, then call search. A free beta API key is offered through the contact page.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/v-modal-vmodal-sdk-android)Star history chart for your README
[](https://hysenlabs.com/projects/v-modal-vmodal-sdk-android)Was this useful?